GTE: a graph learning framework for prediction of T-cell receptors and epitopes binding specificity.

The interaction between T-cell receptors (TCRs) and peptides (epitopes) presented by major histocompatibility complex molecules (MHC) is fundamental to the immune response. Accurate prediction of TCR-epitope interactions is crucial for advancing the understanding of various diseases and their prevention and treatment. Existing methods primarily rely on sequence-based approaches, overlooking the inherent topology structure of TCR-epitope interaction networks. In this study, we present $GTE$, a novel heterogeneous Graph neural network model based on inductive learning to capture the topological structure between TCRs and Epitopes. Furthermore, we address the challenge of constructing negative samples within the graph by proposing a dynamic edge update strategy, enhancing model learning with the nonbinding TCR-epitope pairs. Additionally, to overcome data imbalance, we adapt the Deep AUC Maximization strategy to the graph domain. Extensive experiments are conducted on four public datasets to demonstrate the superiority of exploring underlying topological structures in predicting TCR-epitope interactions, illustrating the benefits of delving into complex molecular networks. The implementation code and data are available at https://github.com/uta-smile/GTE.

Training of epitope-TCR prediction models with healthy donor-derived cancer-specific T cells.

Discovery of epitope-specific T-cell receptors (TCRs) for cancer therapies is a time consuming and expensive procedure that usually requires a large amount of patient cells. To maximize information from and minimize the need of precious samples in cancer research, prediction models have been developed to identify in silico epitope-specific TCRs. In this chapter, we provide a step-by-step protocol to train a prediction model using the user-friendly TCRex webtool for the nearly universal tumor-associated antigen Wilms' tumor 1 (WT1)-specific TCR repertoire. WT1 is a self-antigen overexpressed in numerous solid and hematological malignancies with a high clinical relevance. Training of computational models starts from a list of known epitope-specific TCRs which is often not available for new cancer epitopes. Therefore, we describe a workflow to assemble a training data set consisting of TCR sequences obtained from WT137-45-reactive CD8 T cell clones expanded and sorted from healthy donor peripheral blood mononuclear cells.

Prevention of CMV/EBV reactivation by double-specific T cells in patients after allogeneic stem cell transplantation: results from the randomized phase I/IIa MULTIVIR-01 study.

Introduction: Allogeneic stem cell transplantation is used to cure hematologic malignancies or deficiencies of the hematopoietic system. It is associated with severe immunodeficiency of the host early after transplant and therefore early reactivation of latent herpesviruses such as CMV and EBV within the first 100 days are frequent. Small studies and case series indicated that application of herpes virus specific T cells can control and prevent disease in this patient population. Methods: We report the results of a randomized controlled multi centre phase I/IIa study (MULTIVIR-01) using a newly developed T cell product with specificity for CMV and EBV derived from the allogeneic stem cell grafts used for transplantation. The study aimed at prevention and preemptive treatment of both viruses in patients after allogeneic stem cell transplantation targeting first infusion on day +30. Primary endpoints were acute transfusion reaction and acute-graft versus-host-disease after infusion of activated T cells. Results: Thirty-three patients were screened and 9 patients were treated with a total of 25 doses of the T cell product. We show that central manufacturing can be achieved successfully under study conditions and the product can be applied without major side effects. Overall survival, transplant related mortality, cumulative incidence of graft versus host disease and number of severe adverse events were not different between treatment and control groups. Expansion of CMV/EBV specific T cells was observed in a fraction of patients, but overall there was no difference in virus reactivation. Discussion: Our study results indicate peptide stimulated epitope specific T cells derived from stem cell grafts can be administered safely for prevention and preemptive treatment of reactivation without evidence for induction of acute graft versus host disease. Clinical trial registration: https://clinicaltrials.gov, identifier NCT02227641.

Using imlifidase to elucidate the characteristics and importance of anti-GBM antibodies produced after start of treatment.

BACKGROUND: Autoantibodies are common in glomerulonephritis, but the clinical benefit of rapid elimination has not been determined, even in anti-glomerular basement membrane (GBM) disease. Even less is known about the importance of autoantibody characteristics, including epitope specificity and immunoglobulin G (IgG) subclass distribution. We aimed to address this by characterizing the autoantibody profile in anti-GBM patients: we utilized samples from the GOOD-IDES-01 (treating GOODpasture's disease with Imunoglobulin G Degrading Enzyme of Streptococcus pyogenous) (ClinicalTrials.gov identifier: NCT03157037) trial , where imlifidase, which cleaves all IgG in vivo within hours, was given to 15 anti-GBM patients. METHODS: In the GOOD-IDES-01 trial, plasmapheresis was (re)started if anti-GBM antibodies rebounded. Serum samples were collected prospectively for 6 months and analyzed for anti-GBM epitope specificity using recombinant constructs of the EA and EB epitopes, IgG subclass using monoclonal antibodies, and anti-neutrophil cytoplasmic antibodies (ANCA). The results were correlated with clinical data. RESULTS: Patients with a rebound (n = 10) tended to have lower eGFR at 6 months (11 vs 34 mL/min/1.73 m2, P = .055), and patients with dialysis at 6 months had a higher EB/EA ratio at rebound (0.8 vs 0.5, P = .047). Moreover, two patients demonstrated increasing epitope restriction and several patients displayed a shift in subclass distribution at rebound. Six patients were double positive for ANCA. ANCA rebound was seen in 50% of patients; only one patient remained ANCA positive at 6 months. CONCLUSIONS: In this study, rebound of anti-GBM antibodies, especially if directed against the EB epitope, was associated with a worse outcome. This supports the notion that all means should be used to eliminate anti-GBM antibodies. In this study ANCA was removed early and long-term by imlifidase and cyclophosphamide.

Improved T cell receptor antigen pairing through data-driven filtering of sequencing information from single cells.

Novel single-cell-based technologies hold the promise of matching T cell receptor (TCR) sequences with their cognate peptide-MHC recognition motif in a high-throughput manner. Parallel capture of TCR transcripts and peptide-MHC is enabled through the use of reagents labeled with DNA barcodes. However, analysis and annotation of such single-cell sequencing (SCseq) data are challenged by dropout, random noise, and other technical artifacts that must be carefully handled in the downstream processing steps. We here propose a rational, data-driven method termed ITRAP (improved T cell Receptor Antigen Paring) to deal with these challenges, filtering away likely artifacts, and enable the generation of large sets of TCR-pMHC sequence data with a high degree of specificity and sensitivity, thus outputting the most likely pMHC target per T cell. We have validated this approach across 10 different virus-specific T cell responses in 16 healthy donors. Across these samples, we have identified up to 1494 high-confident TCR-pMHC pairs derived from 4135 single cells.

TEINet: a deep learning framework for prediction of TCR-epitope binding specificity.

The adaptive immune response to foreign antigens is initiated by T-cell receptor (TCR) recognition on the antigens. Recent experimental advances have enabled the generation of a large amount of TCR data and their cognate antigenic targets, allowing machine learning models to predict the binding specificity of TCRs. In this work, we present TEINet, a deep learning framework that utilizes transfer learning to address this prediction problem. TEINet employs two separately pretrained encoders to transform TCR and epitope sequences into numerical vectors, which are subsequently fed into a fully connected neural network to predict their binding specificities. A major challenge for binding specificity prediction is the lack of a unified approach to sampling negative data. Here, we first assess the current negative sampling approaches comprehensively and suggest that the Unified Epitope is the most suitable one. Subsequently, we compare TEINet with three baseline methods and observe that TEINet achieves an average AUROC of 0.760, which outperforms baseline methods by 6.4-26%. Furthermore, we investigate the impacts of the pretraining step and notice that excessive pretraining may lower its transferability to the final prediction task. Our results and analysis show that TEINet can make an accurate prediction using only the TCR sequence (CDR3$\beta $) and the epitope sequence, providing novel insights to understand the interactions between TCRs and epitopes.

TCR-Pred: A new web-application for prediction of epitope and MHC specificity for CDR3 TCR sequences using molecular fragment descriptors.

The search for the relationships between CDR3 TCR sequences and epitopes or MHC types is a challenging task in modern immunology. We propose a new approach to develop the classification models of structure-activity relationships (SAR) using molecular fragment descriptors MNA (Multilevel Neighbourhoods of Atoms) to represent CDR3 TCR sequences and the naïve Bayes classifier algorithm. We have created the freely available TCR-Pred web application (http://way2drug.com/TCR-pred/) to predict the interactions between α chain CDR3 TCR sequences and 116 epitopes or 25 MHC types, as well as the interactions between ß chain CDR3 TCR sequences and 202 epitopes or 28 MHC types. The TCR-Pred web application is based on the data (more 250 000 unique CDR3 TCR sequences) from VDJdb, McPAS-TCR, and IEDB databases and the proposed approach. The average AUC values of the prediction accuracy calculated using a 20-fold cross-validation procedure varies from 0.857 to 0.884. The created web application may be useful in studies related with T-cell profiling based on CDR3 TCR sequences.

Understanding the SARS-CoV-2 Virus Neutralizing Antibody Response: Lessons to Be Learned from HIV and Respiratory Syncytial Virus.

The SARS-CoV-2 pandemic commenced in 2019 and is still ongoing. Neither infection nor vaccination give long-lasting immunity and, here, in an attempt to understand why this might be, we have compared the neutralizing antibody responses to SARS-CoV-2 with those specific for human immunodeficiency virus type 1 (HIV-1) and respiratory syncytial virus (RSV). Currently, most of the antibodies specific for the SARS-CoV-2 S protein map to three broad antigenic sites, all at the distal end of the S trimer (receptor-binding site (RBD), sub-RBD and N-terminal domain), whereas the structurally similar HIV-1 and the RSV F envelope proteins have six antigenic sites. Thus, there may be several antigenic sites on the S trimer that have not yet been identified. The epitope mapping, quantitation and longevity of the SARS-CoV-2 S-protein-specific antibodies produced in response to infection and those elicited by vaccination are now being reported for specific groups of individuals, but much remains to be determined about these aspects of the host-virus interaction. Finally, there is a concern that the SARS-CoV-2 field may be reprising the HIV-1 experience, which, for many years, used a virus for neutralization studies that did not reflect the neutralizability of wild-type HIV-1. For example, the widely used VSV-SARS-CoV-2-S protein pseudotype has 10-fold more S trimers per virion and a different configuration of the trimers compared with the SARS-CoV-2 wild-type virus. Clarity in these areas would help in advancing understanding and aid countermeasures of the SARS-CoV-2 pandemic.

Structural Basis of a Novel Agonistic Anti-OX40 Antibody.

Agonistic antibodies targeting co-stimulating receptor OX40 on T cells are considered as important as (or complementary to) the immune checkpoint blockers in cancer treatment. However, none of these agonistic antibodies have reached the late stage of clinical development partially due to the lack of intrinsic potency with the correlation between binding epitope and activity of the antibody not well understood. Here, we identified a novel anti-OX40 agonistic antibody DF004, which stimulated the proliferation of human CD4+ T cells in vitro and inhibited tumor growth in a mouse model. Our crystallography structural studies showed that DF004 binds to the CRD2 region of OX40 while RG7888, an OX40 agonist antibody developed by Roche, binds to CRD3 of OX40 to the diametrically opposite position of DF004. This suggests that the agonistic activities of the antibodies are not necessarily epitope dependent. As their agonistic activities critically depend on clustering or cross-linking, our structural modeling indicates that the agonistic activity requires the optimal positioning of three Fc receptor/antibody/OX40 complexes on the cell membrane to facilitate the formation of one intracellular hexameric TRAF complex for downstream signal transduction, which is relatively inefficient. This may explain the lack of sufficient potency of these OX40 antibodies in a therapeutic setting and sheds light on the development of cross-linking-independent agonistic antibodies.

Flexible Distance-Based TCR Analysis in Python with tcrdist3.

Paired- and single-chain T cell receptor (TCR) sequencing are now commonly used techniques for interrogating adaptive immune responses. TCRs targeting the same epitope frequently share motifs consisting of critical contact residues. Here we illustrate the key features of tcrdist3, a new Python package for distance-based TCR analysis through a series of three interactive examples. In the first example, we illustrate how tcrdist3 can integrate sequence similarity networks, gene-usage plots, and background-adjusted CDR3 logos to identify TCR sequence features conferring antigen specificity among sets of peptide-MHC-multimer sorted receptors. In the second example, we show how the TCRjoin feature in tcrdist3 can be used to flexibly query receptor sequences of interest against bulk repertoires or libraries of previously annotated TCRs based on matching of similar sequences. In the third example, we show how the TCRdist metric can be leveraged to identify candidate polyclonal receptors under antigenic selection in bulk repertoires based on sequence neighbor enrichment testing, a statistical approach similar to TCRNET and ALICE algorithms, but with added flexibility in how the neighborhood can be defined.

Contribution of T Cell Receptor Alpha and Beta CDR3, MHC Typing, V and J Genes to Peptide Binding Prediction.

Introduction: Predicting the binding specificity of T Cell Receptors (TCR) to MHC-peptide complexes (pMHCs) is essential for the development of repertoire-based biomarkers. This affinity may be affected by different components of the TCR, the peptide, and the MHC allele. Historically, the main element used in TCR-peptide binding prediction was the Complementarity Determining Region 3 (CDR3) of the beta chain. However, recently the contribution of other components, such as the alpha chain and the other V gene CDRs has been suggested. We use a highly accurate novel deep learning-based TCR-peptide binding predictor to assess the contribution of each component to the binding. Methods: We have previously developed ERGO-I (pEptide tcR matchinG predictiOn), a sequence-based T-cell receptor (TCR)-peptide binding predictor that employs natural language processing (NLP) -based methods. We improved it to create ERGO-II by adding the CDR3 alpha segment, the MHC typing, V and J genes, and T cell type (CD4+ or CD8+) as to the predictor. We then estimate the contribution of each component to the prediction. Results and Discussion: ERGO-II provides for the first time high accuracy prediction of TCR-peptide for previously unseen peptides. For most tested peptides and all measures of binding prediction accuracy, the main contribution was from the beta chain CDR3 sequence, followed by the beta chain V and J and the alpha chain, in that order. The MHC allele was the least contributing component. ERGO-II is accessible as a webserver at http://tcr2.cs.biu.ac.il/ and as a standalone code at https://github.com/IdoSpringer/ERGO-II.

Current challenges for unseen-epitope TCR interaction prediction and a new perspective derived from image classification.

The prediction of epitope recognition by T-cell receptors (TCRs) has seen many advancements in recent years, with several methods now available that can predict recognition for a specific set of epitopes. However, the generic case of evaluating all possible TCR-epitope pairs remains challenging, mainly due to the high diversity of the interacting sequences and the limited amount of currently available training data. In this work, we provide an overview of the current state of this unsolved problem. First, we examine appropriate validation strategies to accurately assess the generalization performance of generic TCR-epitope recognition models when applied to both seen and unseen epitopes. In addition, we present a novel feature representation approach, which we call ImRex (interaction map recognition). This approach is based on the pairwise combination of physicochemical properties of the individual amino acids in the CDR3 and epitope sequences, which provides a convolutional neural network with the combined representation of both sequences. Lastly, we highlight various challenges that are specific to TCR-epitope data and that can adversely affect model performance. These include the issue of selecting negative data, the imbalanced epitope distribution of curated TCR-epitope datasets and the potential exchangeability of TCR alpha and beta chains. Our results indicate that while extrapolation to unseen epitopes remains a difficult challenge, ImRex makes this feasible for a subset of epitopes that are not too dissimilar from the training data. We show that appropriate feature engineering methods and rigorous benchmark standards are required to create and validate TCR-epitope predictive models.

T cell receptor sequence clustering and antigen specificity.

There has been increasing interest in the role of T cells and their involvement in cancer, autoimmune and infectious diseases. However, the nature of T cell receptor (TCR) epitope recognition at a repertoire level is not yet fully understood. Due to technological advances a plethora of TCR sequences from a variety of disease and treatment settings has become readily available. Current efforts in TCR specificity analysis focus on identifying characteristics in immune repertoires which can explain or predict disease outcome or progression, or can be used to monitor the efficacy of disease therapy. In this context, clustering of TCRs by sequence to reflect biological similarity, and especially to reflect antigen specificity have become of paramount importance. We review the main TCR sequence clustering methods and the different similarity measures they use, and discuss their performance and possible improvement. We aim to provide guidance for non-specialists who wish to use TCR repertoire sequencing for disease tracking, patient stratification or therapy prediction, and to provide a starting point for those aiming to develop novel techniques for TCR annotation through clustering.

Prediction of Specific TCR-Peptide Binding From Large Dictionaries of TCR-Peptide Pairs.

Current sequencing methods allow for detailed samples of T cell receptors (TCR) repertoires. To determine from a repertoire whether its host had been exposed to a target, computational tools that predict TCR-epitope binding are required. Currents tools are based on conserved motifs and are applied to peptides with many known binding TCRs. We employ new Natural Language Processing (NLP) based methods to predict whether any TCR and peptide bind. We combined large-scale TCR-peptide dictionaries with deep learning methods to produce ERGO (pEptide tcR matchinG predictiOn), a highly specific and generic TCR-peptide binding predictor. A set of standard tests are defined for the performance of peptide-TCR binding, including the detection of TCRs binding to a given peptide/antigen, choosing among a set of candidate peptides for a given TCR and determining whether any pair of TCR-peptide bind. ERGO reaches similar results to state of the art methods in these tests even when not trained specifically for each test. The software implementation and data sets are available at https://github.com/louzounlab/ERGO. ERGO is also available through a webserver at: http://tcr.cs.biu.ac.il/.

Epitope specificity of two anti-morphine monoclonal antibodies: In vitro and in silico studies.

Monoclonal antibodies (mAbs) against morphine are important in the development of immunotherapeutic and diagnostic methods for the treatment and prevention of drug addiction. By the surface plasmon resonance (SPR) and enzyme immunoassay techniques, we characterized two previously obtained mAbs 3K11 and 6G1 and showed their ability to recognize free morphine and morphine-containing antigens in different ways because of the epitope specificity thereof. Using the defined amino acid sequences, we obtained three-dimensional models of the variable regions of Fab fragments of these antibodies and compared them with the known sequence and spatial structure of the anti-morphine antibody 9B1. Docking simulations are performed to obtain models of the antibodies complexes with morphine. Differences in the models of 3K11 and 6G1 complexes with morphine correlate with their experimentally detected epitope specificity. The results, in particular, can be used for the structure-based design of the corresponding humanized antibodies. According to our modeling and docking results, the very different modes of morphine binding to mAbs 3K11 and 6G1 are qualitatively similar to those previously reported for cocaine and two anti-cocaine antibodies. Thus, the obtained structural information brings more insight into the hapten recognition diversity.

Restricted myeloperoxidase epitopes drive the adaptive immune response in MPO-ANCA vasculitis.

BACKGROUND: Treatment of autoimmune diseases has relied on broad immunosuppression. Knowledge of specific interactions between human leukocyte antigen (HLA), the autoantigen, and effector immune cells, provides the foundation for antigen-specific therapies. These studies investigated the role of HLA, specific myeloperoxidase (MPO) epitopes, CD4+ T cells, and ANCA specificity in shaping the immune response in patients with anti-neutrophil cytoplasmic autoantibody (ANCA) vasculitis. METHODS: HLA sequence-based typing identified enriched alleles in our patient population (HLA-DPB1*04:01 and HLA-DRB4*01:01), while in silico and in vitro binding studies confirmed binding between HLA and specific MPO epitopes. Class II tetramers with MPO peptides were utilized to detect autoreactive CD4+ T cells. TCR sequencing was performed to determine the clonality of T cell populations. Longitudinal peptide ELISAs assessed the temporal nature of anti-MPO447-461 antibodies. Solvent accessibility combined with chemical modification determined the buried regions of MPO. RESULTS: We identified a restricted region of MPO that was recognized by both CD4+ T cells and ANCA. The autoreactive T cell population contained CD4+CD25intermediateCD45RO+ memory T cells and secreted IL-17A. T cell receptor (TCR) sequencing demonstrated that autoreactive CD4+ T cells had significantly less TCR diversity when compared to naïve and memory T cells, indicating clonal expansion. The anti-MPO447-461 autoantibody response was detectable at onset of disease in some patients and correlated with disease activity in others. This region of MPO that is targeted by both T cells and antibodies is not accessible to solvent or chemical modification, indicating these epitopes are buried. CONCLUSIONS: These observations reveal interactions between restricted MPO epitopes and the adaptive immune system within ANCA vasculitis that may inform new antigen-specific therapies in autoimmune disease while providing insight into immunopathogenesis.

Detection of Enriched T Cell Epitope Specificity in Full T Cell Receptor Sequence Repertoires.

High-throughput T cell receptor (TCR) sequencing allows the characterization of an individual's TCR repertoire and directly queries their immune state. However, it remains a non-trivial task to couple these sequenced TCRs to their antigenic targets. In this paper, we present a novel strategy to annotate full TCR sequence repertoires with their epitope specificities. The strategy is based on a machine learning algorithm to learn the TCR patterns common to the recognition of a specific epitope. These results are then combined with a statistical analysis to evaluate the occurrence of specific epitope-reactive TCR sequences per epitope in repertoire data. In this manner, we can directly study the capacity of full TCR repertoires to target specific epitopes of the relevant vaccines or pathogens. We demonstrate the usability of this approach on three independent datasets related to vaccine monitoring and infectious disease diagnostics by independently identifying the epitopes that are targeted by the TCR repertoire. The developed method is freely available as a web tool for academic use at tcrex.biodatamining.be.

What has zinc transporter 8 autoimmunity taught us about type 1 diabetes?

Zinc transporter 8 (ZnT8), a protein highly specific to pancreatic insulin-producing beta cells, is vital for the biosynthesis and secretion of insulin. ZnT8 autoantibodies (ZnT8A) are among the most recently discovered and least-characterised islet autoantibodies. In combination with autoantibodies to several other islet antigens, including insulin, ZnT8A help predict risk of future type 1 diabetes. Often, ZnT8A appear later in the pathogenic process leading to type 1 diabetes, suggesting that the antigen is recognised as part of the spreading, rather than the initial, autoimmune response. The development of autoantibodies to different forms of ZnT8 depends on the genotype of an individual for a polymorphic ZnT8 residue. This genetic variant is associated with susceptibility to type 2 but not type 1 diabetes. Levels of ZnT8A often fall rapidly after diagnosis while other islet autoantibodies can persist for many years. In this review, we consider the contribution made by ZnT8 to our understanding of type 1 diabetes over the past decade and what remains to be investigated in future research.

Mapping of Neutralizing Antibody Epitopes on the Envelope of Viruses Obtained from Plasma Samples Exhibiting Broad Cross-Clade Neutralization Potential Against HIV-1.

Several broadly neutralizing antibodies (bNAbs) that can target HIV strains with large degrees of variability have recently been identified. However, efforts to induce synthesis of such bNAbs that can protect against HIV infection have not met with much success. Identification of specific epitopes encoded in the HIV-1 envelope (Env) that can direct the host to synthesize bNAbs remains a challenge. In a previous study, we identified 12 antiretroviral therapy-naive HIV-1-infected individuals whose plasma exhibited broad cross-clade neutralization property against different clades of HIV-1. In this study, we sequenced the full-length HIV-1 gp160 from 11 of these individuals and analyzed the sequences to identify bNAb epitopes. We identified critical residues in the viral envelopes that contribute to the formation of conformational epitopes and possibly induce the production of bNAbs, using in silico methods. We found that many of the sequences had conserved glycans at positions N160 (10/11) and N332 (9/11), which are known to be critical for the binding of PG9/PG16-like and PGT128-like bNAbs, respectively. We also observed conservation of critical glycans at positions N234 and N276 critical for the interaction with CD4 binding site bNAbs in 8/11 and 11/11 sequences, respectively. We modeled the three-dimensional structure of the 11 HIV-1 envelopes and found that though each had structural differences, the critical residues were mostly present on the surface of the Env structures. The identified critical residues are proposed as candidates for further evaluation as bNAb epitopes.

Epitope-specific affinity maturation improved stability of potent protease inhibitory antibodies.

Targeting effectual epitopes is essential for therapeutic antibodies to accomplish their desired biological functions. This study developed a competitive dual color fluorescence-activated cell sorting (FACS) to maturate a matrix metalloprotease 14 (MMP-14) inhibitory antibody. Epitope-specific screening was achieved by selection on MMP-14 during competition with N-terminal domain of tissue inhibitor of metalloproteinase-2 (TIMP-2) (nTIMP-2), a native inhibitor of MMP-14 binding strongly to its catalytic cleft. 3A2 variants with high potency, selectivity, and improved affinity and proteolytic stability were isolated from a random mutagenesis library. Binding kinetics indicated that the affinity improvements were mainly from slower dissociation rates. In vitro degradation tests suggested the isolated variants had half lives 6-11-fold longer than the wt. Inhibition kinetics suggested they were competitive inhibitors which showed excellent selectivity toward MMP-14 over highly homologous MMP-9. Alanine scanning revealed that they bound to the vicinity of MMP-14 catalytic cleft especially residues F204 and F260, suggesting that the desired epitope was maintained during maturation. When converted to immunoglobulin G, B3 showed 5.0 nM binding affinity and 6.5 nM inhibition potency with in vivo half-life of 4.6 days in mice. In addition to protease inhibitory antibodies, the competitive FACS described here can be applied for discovery and engineering biosimilars, and in general for other circumstances where epitope-specific modulation is needed.